The team, including researchers from the universities of Utah and Washington, see their system as a solution to the challenge of evaluating therapists, which has historically relied on the expensive and time consuming method of having a human observer sit in and identify relevant behavior and languages. The paper was published Wednesday in Plos One.
The researchers used the ratings that experts had given to therapists in a batch of previous sessions to train the computer system to identify trends in word usage that are associated with empathetic ratings, or a lack of empathy. Words and phrases that indicate reflective listening show empathy, but instructions or probing words do the opposite.
Once trained the computer system was then fed new recordings of therapy sessions and could effectively evaluate how much empathy the therapist was displaying. If widely deployed, such a system could give therapists immediate feedback on how they’re doing with patients.
The paper focused solely on the ability to measure empathy from word choice. This creates the possibility of a simpler and more economical way to evaluate therapists.
In previous research the team has found other measures of empathy. For example, a high pitch or high vocal energy correlates with low empathy ratings. A lack of synchrony in head motion — whether people are nodding together — is tied to low empathy situations. If a therapist and patient have a similar rhythm to how they speak, the therapist is more empathetic.
In couple’s therapy it’s a bad sign if someone is using the word “you,” and not “I” or “we,” as “you” is tied to blaming and less accountability.
Last week the same researchers published a paper in the journal Interspeech demonstrating that acoustic features in couples therapy could predict the marital outcome better than behavioral descriptions made by humans.
That work could prove useful as a way for therapists to gauge the quality and effectiveness of their counseling session, which would help them in moving couples toward better relationships.
Because the technology is new and still in the lab, it’s not clear yet what features a commercial version of the product would rely on to make judgments.